eduzhai > Applied Sciences > Computer Science >

Exploiting context when learning to classify

  • paiqiu
  • (0) Download
  • 20210805
  • Save

... pages left unread,continue reading

Document pages: 6 pages

Abstract: This paper addresses the problem of classifying observations whenfeatures are context-sensitive, specifically when the testing set involves a contextthat is different from the training set. The paper begins with a precise definition ofthe problem, then general strategies are presented for enhancing the performanceof classification algorithms on this type of problem. These strategies are tested ontwo domains. The first domain is the diagnosis of gas turbine engines. Theproblem is to diagnose a faulty engine in one context, such as warm weather,when the fault has previously been seen only in another context, such as coldweather. The second domain is speech recognition. The problem is to recognizewords spoken by a new speaker, not represented in the training set. For bothdomains, exploiting context results in substantially more accurate classification.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×